Status of Oxide etch rates in Lampoly

Aaron Partridge aaronp at micromachine.stanford.edu
Mon Apr 16 18:56:18 PDT 2001


lam folks,

sandblasting the dielectric sounds reasonable, or perhaps we can find 
a company that will chemical or furnace clean it for us.  i am also 
ok with running O2 cleans and eating the seals -- if that works then 
it seems a viable solution.  i am not enthusiastic about changing the 
etch processes; i would like to keep process #1 as we now have it 
since i have experience with the resultant profiles and etch rates.

one other solution -- cannibalize the alumina plate in lamsi.  this 
might not be as bad as it sounds.  if we ever get lamsi running for 
deep silicon etching it will need a quartz plate anyhow.

aaron.




>Lam Users,
>
>The oxide etch rate have come down a bit but they are still higher than
>they should be. I now feel that the problem is that we have a carbon
>source in the chamber. In CL2 and HBr processes carbon is needed for the
>etching of oxide. The purpose He-O2 flow that we add to the standard poly
>process is to remove carbon during the etch. Normally, 5 sccm of 20% O2 in
>He is suffice to remove carbon coming from the resist mask. However, if
>there is another carbon source in the chamber, 5 sccm may not be enough.
>The problem with adding too much O2 is that the resist etch rate goes up
>with O2 flow. Note that the oxide etch rate normally goes down faster than
>the resist etch rate goes up. About a month ago when I last saw the inside
>of the chamber, I noticed a black buildup on the top dielectric between
>the ICP coil and the chamber. I now suscept this layer is carbon or a
>carbon mixture. The etch rates below support a carbon source in the
>chamber in that the ox etch is coming down as the He-O2 flow is increased.
>We have several choices in what to do about this black layer which are:
>
>1. Do nothing and live with the higher oxide etch rates. O2 plasma clean
>should help but too much O2 cleaning will destory the system o-rings and
>the system plasma.
>
>2. Sand blast the layer off.
>
>3. Turn over dielectric so that the black layer is on the outside toward
>the coil. Note that this dielectric is about an inch thick and 12 inches
>in diameter.
>
>4. Get a new dielectric. We have an old style dielectric in that ours is
>made of aluminia ceramic while Lam currently uses quartz which reduced
>aluminum contamination to the wafers.
>
>Since no one has been complaining about Al contamination in the Lam, my
>choice is to sand blast the dielectric. I would be interested in hearing
>what you think about this.
>
>Here are the results that Brian Green got today using recipe #1.
>
>Program #1 Main Etch only (no BT or OE)
>He-O2 =  0 sccm -> SiO2 etch rate = 220 A/min
>He-O2 =  3 sccm -> SiO2 etch rate = 212 A/min
>He-O2 =  5 sccm -> SiO2 etch rate = 184 A/min <- Normal Program #1 O2 flow
>He-O2 = 10 sccm -> SiO2 etch rate = 167 A/min
>
>Program #1 Over Etch only (no BT or ME)
>He-O2 =  5 sccm -> SiO2 etch rate = 46 A/min <- Normal Program #1 O2 flow
>
>Note: All wafers etched for 2 minutes
>
>I will try to more data as the week goes on.
>
>When we initially set up the Lam, the ox etch rate during the main etch
>was only 35 A/min and during the OE it was 26 A/min. However, we have been
>runnig high for some time.
>
>	Jim McVittie




More information about the lampoly mailing list